CB-EMT: USING CLASSIFICATION AND CLUSTERING TECHNIQUES FOR STUDENT'S PERFORMANCE PREDICTION/ AMMAR ALMASRI; Supervisor: Erbuğ Çelebi, Co-supervisor: Rami Alkhawaldeh

Yazar: Katkıda bulunan(lar):Tanım: p. X, 151; table, figure, chart, 30.5 cm CDİçerik türü:
  • text
Ortam türü:
  • unmediated
Taşıyıcı türü:
  • volume
Konu(lar): Tez notu: Thesis (PhD.) - CYPRUS INTERNATIONAL UNIVERSITY INSTITUTE OF GRADUATE STUDIES AND RESEARCH MANAGEMENT INFORMATION SYSTEM DEPARTMENT Özet: ABSTRACT Currently, in Jordan, and as a result of the increasing volume of students' records, predicting students' performance in academic institutes has become challenging. Most academic institutions need to analyze and monitor the progress of their students using warning prediction systems. The evaluation of students reflects the higher learning reputation of academic institutes. Hence, the performance of students is one of the crucial criteria in establishing the quality of academic institutes. Based on the literature review in chapter 2, the predicting of students' performance rely on two main factors, namely historical students' academic features and prediction model. Accordingly, this thesis is focused on proposing a new dataset and novel prediction model. The first and essential step is focused on evaluating the important features that are used to evaluate students' performance, as shown in chapter 3. In the second phase, which is represented in chapters 4 and 5, focused on proposing a novel prediction model based on baseline classifiers and a coherent dataset. A novel approach named Cluster-Based (CB) classifier that comprises of two phases, clustering and classification was proposed. In the clustering phase, we apply clustering algorithms to group the dataset into a set of clusters. Then, we build a classifier from the records of each group or cluster as a classifier for that group along with a centroid vector. The centroid vectors of the classifiers are used as an indicator for selecting a suitable classifier for the future record of the unknown label. The results showed that the CB-EMT model achieves high performance, and it would assist academic institutes in improving their decision-making, especially for those students who potentially could achieve low-results of marks. This means that the clustering prediction techniques can group the students into homogeneous clusters to help build robust classifiers to predict the students’ CGPA with high evaluation metrics. These results also showed that the proposed CB-EMT model is a promising decision tool for finding potential students with low-performance results who would be the target for many educational institutes to enhance their performance. KEYWORDS: Educational Data Mining (EDM), Ensemble Meta-based Tree (EMT), Clustering Based-EMT (CB-EMT), Feature Selection Methods (FCMs), Student Performance, Machine Learning (ML).
Materyal türü: Thesis
Mevcut
Materyal türü Geçerli Kütüphane Koleksiyon Yer Numarası Durum Notlar İade tarihi Barkod Materyal Ayırtmaları
Thesis Thesis CIU LIBRARY Tez Koleksiyonu Tez Koleksiyonu D 207 A46 2020 (Rafa gözat(Aşağıda açılır)) Kullanılabilir Management Information Systems Department T2058
Toplam ayırtılanlar: 0

Includes CD

Thesis (PhD.) - CYPRUS INTERNATIONAL UNIVERSITY INSTITUTE OF GRADUATE STUDIES AND RESEARCH MANAGEMENT INFORMATION SYSTEM DEPARTMENT

Includes REFERENCES p. 136-151

ABSTRACT
Currently, in Jordan, and as a result of the increasing volume of students' records, predicting students' performance in academic institutes has become challenging. Most academic institutions need to analyze and monitor the progress of their students using warning prediction systems. The evaluation of students reflects the higher learning reputation of academic institutes. Hence, the performance of students is one of the crucial criteria in establishing the quality of academic institutes. Based on the literature review in chapter 2, the predicting of students' performance rely on two main factors, namely historical students' academic features and prediction model. Accordingly, this thesis is focused on proposing a new dataset and novel prediction model. The first and essential step is focused on evaluating the important features that are used to evaluate students' performance, as shown in chapter 3. In the second phase, which is represented in chapters 4 and 5, focused on proposing a novel prediction model based on baseline classifiers and a coherent dataset.
A novel approach named Cluster-Based (CB) classifier that comprises of two phases, clustering and classification was proposed. In the clustering phase, we apply clustering algorithms to group the dataset into a set of clusters. Then, we build a classifier from the records of each group or cluster as a classifier for that group along with a centroid vector. The centroid vectors of the classifiers are used as an indicator for selecting a suitable classifier for the future record of the unknown label. The results showed that the CB-EMT model achieves high performance, and it would assist academic institutes in improving their decision-making, especially for those students who potentially could achieve low-results of marks. This means that the clustering prediction techniques can group the students into homogeneous clusters to help build robust classifiers to predict the students’ CGPA with high evaluation metrics. These results also showed that the proposed CB-EMT model is a promising decision tool for finding potential students with low-performance results who would be the target for many educational institutes to enhance their performance.
KEYWORDS: Educational Data Mining (EDM), Ensemble Meta-based Tree (EMT), Clustering Based-EMT (CB-EMT), Feature Selection Methods (FCMs), Student Performance, Machine Learning (ML).

Araştırmaya Başlarken  
   Sıkça Sorulan Sorular